1. Technical Field
The present invention relates to bonding and more particularly to a bonding method, bonding apparatus and bonding program in which the positions of a plurality of positioning patterns disposed on a chip that is the object of bonding are respectively detected, the positions of bonding pads that are in a specified positional-relationship with the plurality of positioning patterns are calculated, and bonding is performed in the calculated bonding positions.
2. Description of the Related Art
In wire bonding between a plurality of bonding pads disposed on a chip and a plurality of bonding leads disposed on a circuit board, etc., on which the chip is mounted, the bonding of wires is accomplished by moving a bonding tool to the positions of the respective bonding pads and the positions of the respective bonding leads. As chips have become smaller and more highly integrated, the dimensions of bonding pads have become smaller and the spacing of such bonding pads has become narrower; accordingly, the accurate specification of the positions of the respective bonding pads has become necessary. Accordingly, position detection of bonding pads and the positioning patterns used in bonding is practiced.
However, when chips are disposed so that these chips are shifted in the direction of rotation, an inclination is generated in the positioning patterns, etc., so that accurate position detection cannot be accomplished, and wire bonding cannot be performed correctly. Japanese Patent Application Laid-Open (Kokai) No. S63-56764 discloses a method in which pattern matching between a reference-image prepared beforehand and the object-image is performed by successively rotating the reference-image from 0 to 360° for each pattern matching in cases where there is rotation, repeating pattern matching for each angle, and judging the locations and angles that show the best match as a result.
Furthermore, in the method of Japanese Patent No. 2864735, square regions that are to be compared are extracted from object-image signals obtained by imaging, the image signals contained in the extracted square regions are converted into image signals with polar coordinates by way of using the corners of the square regions as the origin, radial-direction patterns for respective specified angles and radial-direction patterns in the reference angles of reference-images prepared beforehand and subjected to a polar coordinate conversion are successively compared, and the comparative angle of the object-image is calculated.
Furthermore, in Japanese Patent Application Laid-Open (Kokai) No. 2002-208010, an approach using a rotation-resistant reference point is disclosed as a means for performing high-precision position detection without performing pattern matching in the rotational direction (which tends to involve an increase in the quantity of calculations) even in cases where the object of comparison is disposed in an attitude that includes positional deviation in the rotational direction. Here, according to Japanese Patent Application Laid-Open (Kokai) No. 2002-208010, the term “rotation-resistant reference point” refers to a point which is such that the error in the position of the object of comparison that is detected in pattern matching of the reference-image and an image of the object of comparison that is obtained by imaging the object of comparison disposed in an attitude that includes positional deviation in the direction of rotation shows a minimum value.
Furthermore, in Japanese Patent Application Laid-Open (Kokai) No. 2002-208010, it is indicated that normalized correlation calculations can be used as one method of pattern matching. Moreover, the following embodiment is indicated as a method for calculating the rotation-resistant reference point.
In the first embodiment, the rotation-resistant reference point is calculated as follows. Specifically, with one corner of the reference-image taken as the center, a rotated image that is rotated +Q° is produced, and the coordinates (X1, Y1) of the point showing the best match as a result of pattern matching between this rotated image and the reference-image are determined. Similarly, a rotated image that is rotated −Q° is produced, and the coordinates (X2, Y2) of the point showing the best match as a result of pattern matching between this rotated image and the reference-image are determined. The coordinates (AX1, AY1) of the rotation-resistant reference point are expressed by the following Equations (1) through (4) using the coordinates (X1, Y1), (X2, Y2) of these two points, the angle Q° and the coordinates (XC1, YC1) of the corner point taken as the center of rotation.AX1=XC1+r·cosα  (1)AY1=YC1+r·sinα  (2)Here, α=tan−1{(X2−X1)/(Y1−Y2)}  (3)r=√{(X2−X1)2+(Y1−Y2)2}/2sinQ  (4)
The rotation-resistant reference point determined by this method is the center of the object in cases where the pattern used is the shape of the object. For example, in the case of a circle, the center point of the circle is the rotation-resistant reference point, and in the case of a square, the center point of the square is the rotation-resistant reference point.
The second embodiment is a simpler method for calculating the rotation-resistant reference point. Specifically, a plurality of rotational center points are set within the reference-image. Then, the reference-image is rotated +Q° about each rotational center point. The amounts of matching between the respective rotated images thus obtained and the reference-image are respectively calculated. Then, a rotational center point with a relatively large amount of matching (among the plurality of rotational center points) is taken as the rotation-resistant reference point. In this case, a rotational center point that is set in the vicinity of the center of the pattern used is taken as the rotation-resistant reference point.
It is indicated that the coordinates of points used in bonding can be determined with high precision, without any need to perform pattern matching in the rotational direction, by thus calculating the coordinates of the rotation-resistant reference point, and taking this point as a bonding alignment point, i.e., a bonding positioning point.
In regard to position detection of positioning patterns, etc., new techniques have been developed as shown below as a demand for increased wire bonding speed has appeared; along with these new techniques, new problems have arisen.
For example, as the scale of LSI has increased, the number of boning pads has increased, and detection of the individual positions of all of these pads takes time. Accordingly, a method is practiced in which only the positions of positioning patterns, at least two of which are spaced as widely as possible on the surface of the chip, are detected, the positions of the other bonding pads are obtained by calculation based upon these positions, and the calculated positions are taken as the bonding target positions.
Furthermore, in cases where wire bonding is performed for numerous types of chips, the storage in memory and read-out of the disposition of positioning patterns and bonding pads on the chip according to the type involved is also bothersome. Accordingly, a method in which a reference chip that acts as a reference for the type of chip involved is prepared when there is a change in the chip type, the disposition relationship of the positioning patterns and bonding pads is stored in memory as training for this reference chip, next, positioning patterns are imaged for the chip that is the object of bonding as the running state, and position detection is accomplished by pattern matching with the acquired image of the reference chip, is also practiced.
If an even greater increase in speed is required, a considerable processing time is required for the pattern recognition of a plurality of positioning patterns in the same visual field; accordingly, a method is performed in which only the areas in the vicinity of the respective positioning patterns are imaged, and positional detection is accomplished by performing pattern matching based upon these images.
Thus, as the speed progressively increases, the question of how to discriminate a plurality of positioning patterns and detect the positions of these patterns in a short time becomes an important performance factor in wire bonding apparatuses.
In this case, it has been demonstrated that if there is a deviation of the chip in the rotational direction, i.e., an inclination, this hinders an increase in the speed of position detection. The conditions of this problem will be described with reference to FIGS. 1 and 2, which show the relationship between the reference chip and the acquired image of the bonding chip. FIG. 1 shows a case in which the bonding object chip 230 is not inclined with respect to the reference chip 200, and FIG. 2 shows a case in which the bonding object chip 230 is inclined with respect to the reference chip 200. In FIGS. 1 and 2, the positioning patterns 202, 212, 232 and 252 are disposed on the upper left corners and lower right corners of the chips 200 and 230.
The training in regard to the reference chip 200 may be described as follows. A first reference-image 204 that includes the first positioning pattern 202 disposed on the upper left corner is acquired, and a second reference-image 214 that includes the second positioning pattern 212 disposed on the lower right corner is similarly acquired. These imaging-positions are detected by the wire bonding apparatus, and are respectively stored in memory as the center position 206 of the first reference-image and the center position 216 of the second reference-image. These center positions are treated as the position of the first positioning pattern 202 and the position of the second positioning pattern 212, and the respective positions of numerous bonding pads (not shown in the drawings) are calculated based upon these positions.
When training is completed, detection of the positioning patterns on the bonding object chip 230 is performed as a running step. First, the bonding object chip 230 is imaged in the position where the first positioning pattern 202 of the reference chip 200 was imaged. As shown in FIG. 1, a portion of the bonding object chip 230 is observed in this imaging range 220. Because of requirements for an increased optical magnification in order to ensure precision and the above-described increase in speed, this imaging range 220 is limited to a narrow range, so that the second positioning pattern 252 of the bonding object chip 230 is not observed. The position of the first positioning pattern 232 of the imaged bonding object chip 230 is shifted with respect to the position 206 of the first positioning pattern 202 of the reference chip 200. The position of the first positioning pattern 232 of the bonding object chip 230 can be determined by pattern matching, by moving the first reference-image 204 in parallel from the position 206, and using the moved position 236 of the first reference-image where this first positioning pattern 202 and the first positioning pattern 232 of the bonding object chip 230 show the greatest amount of overlapping.
Next, the camera must be moved in order to image the second positioning pattern of the bonding object chip 230. Besides the information that was acquired in training, i.e., the position 206 of the first positioning pattern 202 and position 216 of the second positioning pattern 212 of the reference chip 200, and the positions of the respective bonding pads calculated based upon these positions, the information that is obtained in this case is only the position 236 of the first positioning pattern of the bonding object chip 230. Accordingly, as shown in FIG. 1, with the point 236, point 206 and point 216 taken as three of the four points that form a parallelogram, the remaining one point 246 is calculated. This is viewed as the position of the second positioning pattern of the bonding object chip 230, and the center of the imaging range 240 of the camera is moved to this position.
In the case of FIG. 1, since the bonding object chip 230 is not inclined, the movement of the above-described camera involves no problem; the camera movement is performed at a high speed, and imaging of the second positioning pattern of the bonding object chip 230 is performed. If the bonding object chip 230 is inclined as shown in FIG. 2, then the position 256 of the actual second positioning pattern 252 may be located outside of the imaging range 240 of the moved camera. Especially in cases where an attempt is made to increase the speed by narrowing the imaging range, there is a possibility of an increased number of instances in which the second positioning pattern 252 that is to be imaged cannot be captured within the imaging range of the moved camera because of the effects of this inclination. If the second positioning pattern 252 is not captured within the imaging range 240, for example, the operator must perform a search while viewing the visual field of the camera after the apparatus is stopped by a recognition error, and must determine a position that allows the second positioning pattern to be captured. Accordingly, the processing time is greatly increased. If the imaging range is therefore made wider, the processing time required for pattern matching, etc., is increased. Furthermore, in order to broaden the imaging range, it is necessary to reduce the optical magnification, so that the problem of a drop in precision accompanying this reduction of the optical magnification also arises. Furthermore, in cases where a pattern resembling the reference pattern is present within a second visual field, there may be cases in which erroneous detection resulting from pattern matching with this similar pattern occurs. Accordingly, the following problem arises: namely, as a result of this erroneous detection, bonding cannot be performed in the desired pad positions on the chip, so that a defective product is manufactured.
Thus, in cases where there is a deviation of the chip in the direction of rotation, i.e., an inclination, this hinders the increase in the speed of movement to the imaging-position of the next positioning pattern, so that the productivity of the wire bonding apparatus cannot be improved. On the other hand, in the conventional techniques described in the above-described Japanese Patent Application Laid-Open (Kokai) No. S63-56764, Japanese Patent No. 2864735, and Japanese Patent Application Laid-Open (Kokai) No. 2002-208010, inclination-angle detection and position detection are performed with the acquired image as an object. Accordingly, there is no description of the capturing of the positioning pattern in the next imaging-position in any of these techniques.
Furthermore, in Japanese Patent Application Laid-Open (Kokai) No. S63-56764 and Japanese Patent No. 2864735, a detection of the inclination-angle is described; however, there is no description of the relationship between this inclination-angle and the capturing of the positioning pattern in the next imaging-position. In the method in which angle detection is performed using a polar coordinate conversion in Japanese Patent No. 2864735, the precision of inclination-angle detection is greatly influenced by the manner in which the origin is set. For example, in cases where the positioning patterns are circular, polar coordinate development can be performed with good reproducibility if the center of each circle is taken as the origin for polar coordinate development. However, there is no angular dependence of the development pattern, so that angle detection is in fact impossible. In cases where the positioning patterns have an asymmetrical shape, the conditions of the development pattern differ according to the location of the origin of polar coordinate development; as result, the precision of angle detection is affected. Japanese Patent No. 2864735 discloses a method in which respective polar coordinate conversions are performed for the four corners of square regions, and the inclination-angle is determined based upon these conversions. In this case, however, the processing time is long.
Furthermore, in the method of Japanese Patent Application Laid-Open (Kokai) No. 2002-208010, a rotation-resistant reference point is calculated, and this point is used as a bonding positioning point, makes it possible to determine the positions with a high degree of precision. However, the inclination-angles of the positioning patterns cannot be determined.
Thus, in the prior art, in cases where the positioning patterns have an inclination-angle, problems remain in terms of how to move quickly to the next imaging-position.